Gunnar Fant

The FK text-to-speech prosody rules for Swedish are outlined. They cover all levels including prosodic grouping from syntactical analysis. It is a superposition system with local accentuations superimposed on modular F0 patterns of specific rise and decay patterns in successive prosodic groups. F0 in semitones and segmental durations are calculated as a function of lexically determined prominence and position. Speaker normalisation in frequency and time allow the pooling of male and female data in the analysis stage. The main architecture has been successfully tested in French and English synthesis.

One aim of our study is to discuss some of the relations of prosodic parameters to the speech production process, e.g. how speech intensity is related to the vocal tract filter and the voice source and the underlying aerodynamics. A specific problem of phonetic interest is the role of subglottal pressure and fundamental frequency as intensity determinants and their covariation in speech. Our speech analysis displays, incorporating perceptually scaled syllable prominence, are suitable for multilevel studies of speech parameters. A new intensity parameter, SPLH, related to sonority is introduced. In combination with the standard sound pressure level it provides information on the voice source spectral slope. In Swedish, high long stressed vowels approach a semi-closed target, and thus a sonority minimum, which suggests a motor component in prominence perception.